Daily Notes (Posts about mathjax)http://dailynote.simulkade.com/enThu, 29 Jun 2017 20:26:03 GMTNikola (getnikola.com)http://blogs.law.harvard.edu/tech/rssDaily notes - 1 June 2017http://dailynote.simulkade.com/posts/2017-06-01-daily-notes.htmlAA Eftekhari<div><p>I'm in the library and this is my plan for today:</p>
<ul>
<li>Read the recovery data from the data base</li>
<li>Write an objective function based on the Buckley-Leverett formulation</li>
<li>Run the optimization problem in Julia</li>
<li>Optimize the rel-perm parameters</li>
</ul>
<p>Funny story: the paper that has done core flooding and reported the recovery data at high temperature, does not report the viscosity of oil at high temperature. I had to calculate it using this correlation:
$$\ln(\frac{\mu}{\mu_0}) = B(\frac{1}{T}-\frac{1}{T_0})$$
I found the B value by fitting the above equation to the viscosity data of n-Dodecane. The viscosity data comes from the cool <a href="http://www.coolprop.org">CoolProp</a> package. </p>
<p>The recovery data is not a straight line at the beginning of the core flooding. It must be a problem in choosing the right moment to start the timer. I shifted the curve to make it consistent. </p>
<p>The most important problem with using the analytical solution of the BL equation as an objective function is that very occasionally, the shock front cannot be found with a reasonable numerical accuracy. This blows up the objective function and the whole optimization procedure. Perhaps, I should try the numerical solution (upwind) that is slower, but (almost) always converges. </p></div>mathjaxhttp://dailynote.simulkade.com/posts/2017-06-01-daily-notes.htmlThu, 01 Jun 2017 06:34:44 GMT